A Short Breast Imaging Reporting and Data System-Based Description for Classification of Breast Mass Grade

Area of Science:

• Medical Imaging
• Computational Pathology
• Artificial Intelligence in Healthcare

Background:

• Early breast cancer detection relies on accurate identification of breast masses.
• Current computational methods often focus on general benign/malignant classification, not specific risk grades.
• The Breast Imaging Reporting and Data System (BI-RADS) standard categorizes mass risk based on shape, margin, and density.

Purpose of the Study:

• To develop and test a methodology for classifying breast masses based on BI-RADS descriptors.
• To identify image processing descriptors that correlate with clinical assessments of mass risk.
• To combine neural networks and image processing for automated BI-RADS classification.

Main Methods:

• Utilized the INbreast dataset for analysis.
• Tested various image processing descriptors to find those relevant to clinical assessment.
• Employed a combination of neural networks and image processing techniques.
• Classified masses according to BI-RADS grades (BI-RADS-2 to BI-RADS-5).

Main Results:

• Achieved a general accuracy and sensitivity of 0.88±0.07 in classifying masses.
• Successfully identified masses associated with BI-RADS grades 2 through 5.
• Demonstrated the potential for automated classification linked to clinical practice.

Conclusions:

• The proposed methodology shows promise for automated breast mass classification aligned with clinical standards.
• This approach can assist medical experts by providing descriptions directly related to BI-RADS assessment.
• Further validation on diverse datasets is warranted to generalize the findings.